Short-Term Solar Irradiance Prediction Based on Multichannel LSTM Neural Networks Using Edge-Based IoT System

Table 2

Multifeature input model prediction result evaluation index.

	10 min					30 min					60 min
	MAE	RMSE		SMAPE	MAPE	MAE	RMSE		SMAPE	MAPE	MAE	RMSE		SMAPE	MAPE
Bagging	29.33	68.9	0.942	12.42	17.04	42	84.42	0.9	20.62	36	48.63	90.83	0.9	23.8	75.1
MLP	28.28	65.4	0.948	15.82	24.92	41	79.91	0.91	24.46	53	47.47	88.36	0.9	24.4	62.2
LSTM	33.04	68.3	0.943	18.61	24.89	51	83.77	0.902	25.02	47	54.31	92.63	0.89	26.3	77.27
BiLSTM	34.14	68.3	0.944	19.06	28.67	46	81.04	0.908	25.85	47	56.99	90.91	0.89	27.4	57.34
CNN-LSTM	32.58	67.7	0.944	15.15	22.72	44	78.48	0.914	23.38	45	57.62	89.98	0.9	27	48.82
CNN-BiLSTM	30.11	66.6	0.946	15.14	18.72	46	79.82	0.911	25.61	42	48.04	82.74	0.91	26	49.69
WT-LSTM	11.67	15.8	0.996	6.72	9.62	18	24.61	0.991	12.32	16	23.2	33.76	0.99	13.3	17.25
WT-BiLSTM	11.26	15.7	0.996	8.12	10.51	17	22.45	0.992	15.49	22	25.05	34.08	0.99	12.1	24.16
Proposed	7.66	10.5	0.998	6.13	8.54	9.2	13.98	0.997	8.71	9.5	18.13	27.98	0.99	10.97	15.63